CrawLLM is a dungeon crawler game featuring card-based combat, where the content (cards, level, textures, enemies, and backstory) is generated every time in a top-down fashion via Large Language Models (Mixtral 8x7B) and Large Media Models (Stable Diffusion XL). Generation is guided and controlled by the existing game design (with game progression and cards pre-authored by a designer) while graphical assets are constrained via ControlNet and animation guidelines.

Relevant Publications

  • Marvin Zammit, Antonios Liapis, Georgios N. Yannakakis: "CrawLLM: Theming Games with Large Language Models," in Proceedings of the IEEE Conference on Games, 2024. PDF BibTex